Manufacturing & Quality Control

Vantage connects to shop-floor databases (MES, SCADA historians, ERP) and applies AI-powered analysis to production data — enabling manufacturers to detect quality drift in real time, trace defects back to root causes, and automate compliance documentation without custom software.

Predict Quality Issues Before They Hit the Line

Ingest machine sensor data, detect parameter drift before it causes defects, and alert process engineers automatically.

Scenario: A discrete manufacturer running three shifts needs to monitor temperature, vibration, and cycle time across 40 CNC work cells and detect anomalies before they produce scrap.

Workflow Steps:

1. Schedule Trigger — Run every 15 minutes during active shifts
2. Database Query (PostgreSQL) — Pull real-time sensor data from the SCADA historian: machine ID, temperature (°C), vibration (mm/s RMS), cycle time (seconds), and part count since last tool change
3. Database Query (MSSQL) — Pull current production order details from the ERP: part number, material lot, customer, quantity remaining
4. Join — Merge sensor readings with production context on machine ID
5. Aggregation — Calculate rolling 1-hour statistics per work cell: mean temperature, max vibration, standard deviation of cycle time, and parts-per-hour throughput
6. Computed Column — Calculate control chart indicators:
   • Temperature deviation from setpoint: abs(actual_temp - target_temp)
   • Vibration Z-score: (current_vibration - rolling_mean) / rolling_stddev
   • Cycle time Cp/Cpk: process capability index
7. AI Enrichment — Classify each machine's current state:
   • NORMAL: all parameters within 2σ of control limits
   • DRIFT: one or more parameters trending toward limits (predict time to exceedance)
   • ANOMALY: parameter outside 3σ or sudden step change
   • Include probable root cause: tool wear, material variation, coolant temperature, fixture misalignment
8. Multi-Conditional — Route by classification:
   • ANOMALY → Send Email to plant manager + Send Message (Slack #quality-alerts) with machine ID, parameter, and root cause hypothesis + Create Issue (Jira) for immediate investigation + Dashboard Output (Event Monitor Tile flashes red)
   • DRIFT → Send Message (Slack) to process engineering channel with drift direction and predicted time-to-exceedance + Dashboard Output (Line Tile highlights trending parameter)
   • NORMAL → Dashboard Output only
9. Dashboard Output — Populate:
   • Map Tile — Shop floor layout with machines color-coded: green (normal), yellow (drift), red (anomaly)
   • Line Tile — Real-time parameter trends per machine (temperature, vibration, cycle time) with control limits overlaid
   • Metric Tile — OEE (Overall Equipment Effectiveness), first-pass yield %, scrap rate
   • Gantt Tile — Production schedule: current order, next order, estimated completion
   • Predictive Insights Tile — AI-predicted failure probabilities by machine for the next 24 hours
   • Stat Tile — Parts produced this shift, target vs. actual
10. Write Excel — Archive shift-level quality summary for regulatory traceability (ISO 9001, IATF 16949)

Key Nodes: Schedule Trigger, Database Query (PostgreSQL, MSSQL), Join, Aggregation, Computed Column, AI Enrichment, Multi-Conditional, Write Excel, Dashboard Output, Send Email, Send Message, Create Issue (Jira)

Trace Batch Genealogy from Raw Material to Finished Good

Map the full material genealogy tree when a defect is discovered — from raw material supplier through sub-assembly to finished product and customer shipment.

Scenario: A food manufacturing company discovers contamination in a finished product. They need to trace the affected material lot backward to the supplier and forward to every downstream product and customer shipment within minutes — not hours.

Workflow Steps:

1. Logical Trigger — Fire when a non-conformance report (NCR) is logged against a lot number
2. Database Query (PostgreSQL) — Retrieve the flagged lot: material description, supplier, receiving date, receiving inspection results
3. Database Query (MSSQL) — Pull the bill of materials (BOM) explosion: which sub-assemblies and finished goods used this lot
4. Build Adjacency Maps — Construct the full genealogy tree:
   • Upstream: Supplier → PO → Receiving Lot → Inspection Result
   • Downstream: Raw Material → Sub-Assembly Lot → Finished Good Lot → Work Order → Shipment → Customer
5. Filter — Isolate all downstream finished goods and shipments that contain the affected material
6. Aggregation — Summarize impact:
   • Number of affected finished good lots
   • Number of affected shipments
   • Number of affected customers
   • Total units at risk
7. Geocode — Resolve customer and warehouse addresses to coordinates
8. Multi-Conditional — Route by containment scope:
   • Product still in-plant → Create Issue (Jira) for quarantine hold + Send Message (Slack #quality)
   • Product shipped to warehouse → Send Email to logistics team with hold instructions + Dashboard Output
   • Product delivered to customer → Send Email to quality VP + Send Email to affected customers (recall notification draft) + Dashboard Output (critical alert)
9. Dashboard Output — Populate:
   • Map Tile — Affected facility and customer locations
   • Table Tile — All affected lots with status: quarantined, in-transit, delivered
   • Gantt Tile — Timeline: material receipt → production → shipment → delivery
   • Event Feed Tile — Non-conformance log with containment actions
10. Write PDF — Generate a containment report for FDA/USDA or customer audit
11. Write CSV — Export affected lot listing for ERP update

Key Nodes: Logical Trigger, Database Query, Build Adjacency Maps, Filter, Aggregation, Geocode, Multi-Conditional, Create Issue (Jira), Write PDF, Write CSV, Dashboard Output, Send Email, Send Message

Track OEE and Production Performance in Real Time

Calculate and track Overall Equipment Effectiveness in real time across all production lines.

Scenario: A plant manager wants to see OEE by line and by shift, identify the biggest losses (downtime, speed loss, quality loss), and track improvement initiatives.

Workflow Steps:

1. Schedule Trigger — Run every 30 minutes
2. Database Query (PostgreSQL) — Pull machine state data: running, idle, down, setup time by machine and shift
3. Database Query (MSSQL) — Pull production counts: good parts, rejected parts, expected cycle time
4. Join — Merge state data with production counts on machine ID and time window
5. Computed Column — Calculate OEE components:
   • Availability = (scheduled_time - downtime) / scheduled_time
   • Performance = (actual_cycle_time * total_parts) / available_time
   • Quality = good_parts / total_parts
   • OEE = Availability * Performance * Quality * 100
6. Aggregation — Roll up by production line, shift, and plant
7. Sort — Lines with lowest OEE first (focus improvement on bottlenecks)
8. Filter — Lines with OEE < 65% (world-class is 85%+)
9. AI Enrichment — Classify the dominant loss category for each underperforming line and suggest improvement actions: "Line 3 OEE is 58%. Availability is the primary loss (62%) due to unplanned downtime averaging 47 minutes per shift. The top downtime reason code is 'tool change' — consider implementing SMED methodology."
10. Dashboard Output — Populate:
    • Bar Tile — OEE by production line (stacked: availability, performance, quality)
    • Comparison Tile — This week vs. last week OEE by line
    • Line Tile — OEE trend over 30 days
    • Pivot Tile — OEE matrix: Line × Shift
    • Pie Tile — Downtime reasons (top 5 Pareto)
    • Metric Tile — Plant-wide OEE with sparkline
    • Forecast Tile — Projected OEE if current improvement trajectory continues
11. Send Message (Teams) — Post shift summary to the operations channel: "Shift 2 OEE: 74.2%. Top loss: Line 5 speed loss (Performance: 68%)."

Key Nodes: Schedule Trigger, Database Query (PostgreSQL, MSSQL), Join, Computed Column, Aggregation, Sort, Filter, AI Enrichment, Dashboard Output, Send Message

Manage Supplier Quality with Automated Scorecarding

Track incoming material quality by supplier, identify trends, and automate corrective action requests.

Scenario: A procurement quality team receives materials from 50+ suppliers and needs to track incoming inspection results, calculate supplier scorecards, and trigger corrective action requests when quality drops.

Workflow Steps:

1. Schedule Trigger — Run daily
2. Database Query (PostgreSQL) — Pull incoming inspection records: supplier, material, lot number, inspection date, pass/fail, defect type, defect count
3. Aggregation — Calculate supplier quality metrics over the past 90 days:
   • Lot acceptance rate: passed_lots / total_lots * 100
   • Defects per million (DPMO): (defect_count / total_units_inspected) * 1,000,000
   • On-time delivery rate
   • Average lead time variance
4. Computed Column — Calculate composite supplier score: weighted average of quality (40%), delivery (30%), cost (20%), and responsiveness (10%)
5. Sort — Rank suppliers by score
6. Filter — Suppliers with acceptance rate < 95% or DPMO > 1000
7. Multi-Conditional — Route by severity:
   • Acceptance rate < 90% or DPMO > 5000 → Send Email to supplier (formal corrective action request) + Send Email to procurement director + Create Issue (Jira) for supplier quality investigation
   • Acceptance rate 90–95% → Send Email to supplier (quality concern notice) + Dashboard Output to watchlist
   • Acceptance rate ≥ 95% → Dashboard Output only (recognized as preferred supplier)
8. Write PDF — Generate monthly supplier quality scorecard (one per supplier)
9. Write Excel — Generate consolidated supplier comparison workbook
10. Dashboard Output — Populate:
    • Table Tile — Supplier scorecard: name, acceptance rate, DPMO, delivery rate, composite score
    • Scatter Tile — Quality (acceptance rate) vs. Cost per unit (identifies high-cost/low-quality outliers)
    • Line Tile — Supplier quality trends over 12 months
    • Histogram Tile — Distribution of supplier scores
    • Metric Tile — Average supplier quality score, % of suppliers above target

Key Nodes: Schedule Trigger, Database Query, Aggregation, Computed Column, Sort, Filter, Multi-Conditional, Write PDF, Write Excel, Send Email, Create Issue (Jira), Dashboard Output

Plan Maintenance and Route Work Orders Automatically

Schedule preventive maintenance based on equipment run hours and condition data, and track work order completion.

Scenario: A maintenance department manages 200+ pieces of equipment. They need automated PM scheduling based on run hours, condition monitoring triggers, and full visibility into work order backlog.

Workflow Steps:

1. Schedule Trigger — Run daily at 5 AM (before first shift)
2. Database Query (PostgreSQL) — Pull equipment master: asset ID, description, location, last PM date, PM interval (hours), current run hours since last PM
3. Database Query (MSSQL) — Pull condition monitoring data: vibration trends, oil analysis results, temperature readings
4. Computed Column — Calculate:
   • Hours until next PM: pm_interval - hours_since_last_pm
   • PM compliance: on-time_completions / total_PMs * 100
5. Filter — Equipment due for PM within 7 days or with condition indicators exceeding maintenance thresholds
6. AI Enrichment — Prioritize the maintenance queue: "Asset CNC-042 should be prioritized: vibration trending up 15% over 30 days while PM is due in 3 days. Risk of unplanned failure: 23%."
7. Multi-Conditional — Route by urgency:
   • Overdue PM → Create Issue (Jira, Priority: High) + Send Email to maintenance supervisor
   • Due within 3 days → Create Issue (Jira, Priority: Medium) + Send Message (Slack)
   • Due within 7 days → Create Issue (Jira, Priority: Low)
   • Condition-based alert → Create Issue (Jira) with condition data + Send Email to reliability engineer
8. Dashboard Output — Populate:
   • Gantt Tile — Maintenance schedule: upcoming PMs by asset and date
   • Table Tile — Work order backlog with priority, status, and assigned technician
   • Metric Tile — PM compliance %, overdue count, MTTR
   • Bar Tile — Work orders by type (preventive, corrective, condition-based)
   • Predictive Insights Tile — Predicted unplanned downtime based on condition trends

Key Nodes: Schedule Trigger, Database Query, Computed Column, Filter, AI Enrichment, Multi-Conditional, Create Issue (Jira), Send Email, Send Message, Dashboard Output

Example Dashboard: Shop Floor Operations Center

Build this dashboard to give your plant management real-time visibility into production performance, quality, and maintenance across all lines.

Row 1 — Plant Performance

Row 2 — Machine Health & Schedule

Row 3 — Quality & Trends

Row 4 — Supplier & Maintenance

Row 5 — Traceability & Insights

Getting Started

To build manufacturing workflows:

1. Connect your shop-floor databases — Add your SCADA historian, MES, and ERP databases under Integrations (PostgreSQL, MSSQL, or MongoDB)
2. Start with OEE — Build a simple Schedule Trigger → Database Query → Computed Column → Dashboard Output workflow for OEE tracking
3. Add quality monitoring — Layer in AI Enrichment to classify machine state and detect drift
4. Build traceability — Use Build Adjacency Maps to map material genealogy
5. Automate alerts — Multi-Conditional to route anomalies to the right people via Slack, email, and Jira